CPU architecture - what does this really mean

[jbourne84]

I am beginning to shop around for a cpu to build a new system off of, looking at both AMD and Intel and am trying to get my bearings for where the value really is in each companies lineup.

My last build I went straight for Intel but now that Im trying to compare the two and since AMD seems to be priced lower its tough to know how to compare them at first glance.

one thing I have heard people talk about is how efficient the CPU architecture is, implying that two similarly spec'd CPUs could differ in how efficient they are which is why prices might vary even though they seem very similar.

is there any truth to this? and how can you look for this as a factor when shopping around?(other than assuming it with price variation)

for example -

AMD Phenom II X4 975 - 3.6ghz quad core - around 160.00
Intel Core i5-2500K Sandy Bridge - 3.3GHz (3.7GHz Turbo Boost) - around 230.00

and I think the Intel even has a smaller cache

 

[majorgibly]

I am beginning to shop around for a cpu to build a new system off of, looking at both AMD and Intel and am trying to get my bearings for where the value really is in each companies lineup.

My last build I went straight for Intel but now that Im trying to compare the two and since AMD seems to be priced lower its tough to know how to compare them at first glance.

one thing I have heard people talk about is how efficient the CPU architecture is, implying that two similarly spec'd CPUs could differ in how efficient they are which is why prices might vary even though they seem very similar.

is there any truth to this? and how can you look for this as a factor when shopping around?(other than assuming it with price variation)

for example -

AMD Phenom II X4 975 - 3.6ghz quad core - around 160.00
Intel Core i5-2500K Sandy Bridge - 3.3GHz (3.7GHz Turbo Boost) - around 230.00

and I think the Intel even has a smaller cache

I know this does not answer your question directly but I believe the 2500k is better than the 975. I'm sure someone else can tell you why. I thought it had something to do with nanometers like 32nm...someone else will fill in the gaps.

 

[Goldengoose]

I can't answer your question on why, but i can show you benchmarks i found in another thread on here of the two compared to show the i5 is defiantly better:
http://www.anandtech.com/bench/Product/88?vs=288
Just scroll down.

 

[jemm]

CPU architecture is the layout of the cpu, it is its design -- where the cores, caches, etc are placed inside the cpu. Yes, it is important when buying a PC.

AMD Phenom II X4 975 has a different CPU architecture than Intel Core i5-2500K Sandy Bridge. However the most important thing is what the PC you want will be used for, for instance, gaming, editing/rendering, etc.

As soon as you decide the use of the PC, then you can research on the best parts. It is a good idea to post your inquire here as people are more than qualified to make recommendations.

 

[vegettonox]

Think of it this way, a crowded warehouse, every time the supervisor comes into the building he improves policy and how goods are handled. Essentially every new processor that comes out is like an enhanced system for handling the materials, some are more efficient than others. At the moment the sandy bridge and ivy bridge processor lines from intel are generally better than the amd counterparts but there are some discrepancies.

The 975 processor you are looking at is priced at roughly its performance level, you would be getting a greater value out of the 2500 processor than you would out of the 975, its worth the extra bucks.

Here is a direct comparison between the two processors done by tomshardware themselves =on&prod[4951]=on]http://www.tomshardware.com/charts/desktop-cpu-charts-2010/compare,2417.html?prod[4788]=on&prod[4951]=on

 

[Houndsteeth]

The Intel SB processors have a higher IPC than the PhII processors. What this means is that cycle for cycle, each Core i5 core is able to perform about 1/3 more operations than each PhII core. So, in this case, raw Ghz doesn't really enter into it.

<Ubiquitous car analogy> It's like comparing a car with a bus when you are talking about the ability to carry people. Sure, the car runs faster, but at most, it only carries four passengers per trip while the bus carries 40 passengers. The car has to make 10 trips before it is able to match what the bus does in one trip. Even if the car is able to move at 60 kph vs. the bus at 45 kph, the bus still wins. </Ubiquitous car analogy>

In the case of the Phenom II vs the Core i5, it's more like having an extra row of seats in the back of a car that is able to drive at nearly the same speed.

 

[vitornob]

Another architecture 'tale':

Person 1 - Run at 10mph (this is the clock)
Person 2 - Run at 12mph (this is the clock)

Task: Buy items, in different places. (this is the software requisition)

Person 1 - Take the task, make a map with an optimized path. Path = 10 miles
Person 2 - Take the task, make a map with a not so optimized path. Path = 18 miles

Person 1 - Needs 1 hour
Person 2 - Needs 1 and half hours.

Person 1 is slower (clock) but do the tasks faster.

This is a simplified view of how an 'architecture' makes a 'slower' CPU being faster.

 

[13thmonkey]

The architecture is of academic interest, to someone buying a new system performance and cost are the two most important factors (leaving out any references to ethics, loyalty etc.)

 

[Blandge]

The CPU Microarchitecture refers to all of the design decisions made by the engineers that develop a piece of hardware. This includes how the cache works, the component and bus layout, pipelining decisions etc. The architecture does not generally designate device (transistor level) properties, though most architectures are designed with a specific process technology in mind.

To put it simply, the CPU Architecture is a desciption of all components in the processor and how they interact with each other to make the CPU function. Intel's microarchitecture is vastly different from AMD's even though their general performance is described in the same way (Frequency, Bandwidth, CPI).

13thmonkey is correct is saying that understanding the microarchitecture is unimportant from a consumer standpoint. The important metrics are cost/performance/power/reliability under intended workloads. These are most easily calculated by running synthetic benchmarks, but benchmarking realworld software is more reliable in many cases.

To answer your question about efficiency, you have to specify what type of efficiency. You are most likely interested in performance per clock, which is measured by CPI (Cycle per instruction) or IPC (Instructions per cycle) which describes the amount of work that a computer does per unit of frequency (GHz). In general Intel's processors have a higher CPI and therefore can do more work than AMD's chips at the same frequency. (IE, @ 1GHz each, an Intel chip will get a higher benchmark score than AMD's) This is not ALWAYS the case as we saw with P4 vs. K8, but current Intel chips will have better efficiency in both performance/clock and performance/watt.

As far as performance/dollar that varies greatly from CPU to CPU and workload to workload. It really depends what you want to do. In general, performance/dollar goes down as performance goes up, so to get the CPU with the best performance/dollar will mean your CPU may underperform your expectations. It's best to set a performance goal, and choose the cheapest CPU that meets that minimum performance.

If you are interested in computers I would highly suggest reading some books on the subject.

 

[13thmonkey]

The CPU Microarchitecture refers to all of the design decisions made by the engineers that develop a piece of hardware. This includes how the cache works, the component and bus layout, pipelining decisions etc. The architecture does not generally designate device (transistor level) properties, though most architectures are designed with a specific process technology in mind.

To put it simply, the CPU Architecture is a desciption of all components in the processor and how they interact with each other to make the CPU function. Intel's microarchitecture is vastly different from AMD's even though their general performance is described in the same way (Frequency, Bandwidth, CPI).

13thmonkey is correct is saying that understanding the microarchitecture is unimportant from a consumer standpoint. The important metrics are cost/performance/power/reliability under intended workloads. These are most easily calculated by running synthetic benchmarks, but benchmarking realworld software is more reliable in many cases.

If you are interested in computers I would highly suggest reading some books on the subject.

A very good considered reply, welcome to the forums.

 

[jbourne84]

thanks for the in depth reply, that definitely answered my question.

as it has been said, that actual architecture doesn't really concern me (the consumer) so much as how it affects efficiency for what I want to do, which it sounds like comparing benchmarks is the only way to determine this reasonably.

For the record, the machine is going to be for autoCAD and some 3D modeling (work), but mostly picture and music editing (hobbies) and then maybe some gaming if i have time.

So now my question is how to best compare benchmark tests, the site given earlier, anandtech.com, gave a great breakdown with like 30 different comparisons for those two CPUs.

Is this a good go-to site for this or does anyone have other recommendations?

 

[13thmonkey]

anand is good, but anyone with tests that match your needs would be great